Will AVs [1] mean more or fewer household vehicles? Will they generate more VMT [2] or less? Promote more or less PMT [3]? Opinions range widely from “AVs will be clean and generally shared so no one would want to own one” to “AVs will be so useful and desirable that VMT and PMT will both rise dramatically as household ownership will remain the dominant paradigm.”

In this chapter we consider the potential reasons and links of causality for a rise or fall in relative, worldwide PMT as the robotic-driver triumphs over the human-driver.

Before we review lists of PMT stimulators or dampeners in an AV context, consider several non-technical, socio-economic forces already in place that will influence PMT regardless of automation. These will mostly likely continue to push PMT demand upward, but will certainly be sustaining at a regional level. Regardless of any new technology disruptions or interactions, these form the basis on which automobility will play out as the AV moves from persuasive to pervasive. We will not be able to pin the eventual 2005-2045 PMT differential entirely on the AV. Even this foundational basis will be a developing process—adding to unknowability.

Sociobiology. Since the dawn of civilization 5,000-7,000 years ago humans sought powered automobility. Our need to use powered assistance to travel and transport is a fundamental human sociobiological trait [Grus-13]. It always gets satisfied in some way, whether donkey, camel or cart. Cars and trucks are the revealed preference in the developed world and most PMT is from owning and driving a personal vehicle (see figure). Motorized vehicles are also the stated preference in most of the world, even in the countries with populations that have little current access to them. This creates powerful upward pressure for PMT, and shows no sign of abating.

 

2009 NHTS US Surface PMT (non-freight)

This data from the 2009 National Household Travel Survey illustrates the mobility choices currently made in the US, one of a handful of countries that have already reached automotive saturation. Other countries are likely to tend toward this skewed distribution of revealed preferences as they rely increasingly on motorized vehicles.

Wealth. One knowable link is how far wealth acquisition leverages this sociobiological tendency: car ownership tends to saturate at 0.7-0.8 as populations get wealthier [Darg-07]. We can reasonably estimate, then, that as the human population plateaus at 10 or 11 billion and as that population tends toward a level of wealth that affords more powered automobility (both are predicted), we would be closing in on 7.5 billion vehicles worldwide around 2100 (up from 1.2 billion in 2015). Hence automotive saturation will mature and spread worldwide from the handful of countries such as Australia, UK, and US where it began saturating earlier in the century. This, in combination with human sociobiological traits, implies PMT growth will not abate until saturation in the next century.

At our current clip, we will be at four billion vehicles by mid century—i.e., easily halfway to the expected mature complement of 7.5 billion by 2050.

Even if robotic vehicle fleets were to effectively extinguish personal ownership, we can still expect to experience a shared PMT demand equivalent to that from the otherwise projected level of ownership—PMT demand will be satisfied either way. Given AVs, PMT demand will tell us more about shared-fleet sizing than about ownership, since shared-fleets can be optimized while ownership cannot.

Marchetti distance. Long before civilization, long before humans clustered into cities, our species fell into patterns of travel that averaged around one hour per day [Marc-94]. This daily hour provided a travel distance that limited the radius of human settlement patterns. Today, faster vehicles increase our travel distance so that, and our “Marchetti radius” has expanded At the same time, congestion tends to shrink it. This time budget stays constant making it unlikely humans would consume more or less than this average in daily travel time. If robotics were to reduce congestion or increase travel speed, as is often claimed, PMT would tend to increase since that Marchetti hour remains available. It is safe to conclude that AV PMT will not shrink due to the Marchetti distance constraint.

Marchetti budget. Marchetti also observed that humans spend about 12% of their disposable wealth on travel. An increase in what a Marchetti budget can buy does not mean more travel hours, rather that humans would elect to travel using faster, more convenient or more comfortable means. If the average cost of a mile in an AV will be lower than the per-mile cost of current vehicles, then the Marchetti budget can buy faster, better or more convenient PMT, hence PMT would rise, compounding the PMT effect of worldwide growth in wealth.

These four baseline factors: sociobiological foundation of automobility, wealth and automotive ownership, and the two Marchetti ratios are deeply embedded human social behaviors that form part of our complex relationship with automobility. As Wendell Cox so aptly put it:

[Automobility] “is not based on the shallow, arbitrary preference expressed in the threadbare cliché of a ‘love affair with the automobile.’ Cars are essential to realizing the aspirations of a majority of people, not only in the United States but in Europe and beyond.” [Cox-15]

None of these four factors are downward indicators. All four point to sustained or increased PMT. Given this immutable basis of human behavior, what might make PMT cool off given the introduction of robotics—or heat up even more aggressively?

One last thing before we explore this question. There are three common measures of automobile use: VMT (vehicle miles traveled), PMT (person miles traveled) and vehicle ownership rates (vehicles per 1000 population or per household). VMT is positively correlated with congestion and emissions among several other effects. PMT is usually negatively correlated—at least with respect to congestion and emissions. Since not owning a car given current technology circumstances means that a person drives less, reducing ownership rates implies reduced VMT and to a lesser degree reduced PMT—at least with respect to private vehicles.

In a future dominated by autonomous vehicles, robotics have the potential to alter these correlations considerably. To talk about these changing correlations and their effects, we need to add an additional definition: EMT or empty miles traveled. We already have the equivalent of EMT, but this is seldom underscored. When a taxi is cruising for a fare or when someone is on the way to pick up a child from school, those parts of the trip come about only so that the real purpose of the trip can commence. After the advent of the autonomous vehicle, EMT will be far more apparent due to deadheading. Worse, one might expect an increase in EMT—perhaps a substantial one, since the human time-cost of an EMT goes to zero.

We think it is reasonable to expect massive urban fleets of shared vehicles averaging four or more times the annual VMT and/or PMT per vehicle than consumed today. The several simulations we discussed in Chapter β showed this consistently—usually from seven to 10-fold increases. This means that there is an evident opportunity to transform the relationship between vehicle population and VMT. If the average 2045 motor vehicle provides four times the PMT of a 2015 vehicle, what happens to VMT? It would go up with deadheading and down with parallel sharing (ride-sharing). Since we expect to see only a modest degree of ride-sharing (Chapter β), EMT could become a dominant barrier to the secondary goal of reducing VMT and congestion. To count these miles in a way to make meaningful comparisons between the pre- and post-AV eras, we need to count each EMT as one PMT and one VMT—which is what we essentially do now for the deadheading portion of a chauffeuring trip. It is just that an AV EMT will take on a different Marchetti value than it does pre-AV. It is this that leads to the fear that future AV owners might send their vehicle circling the block while they shop in order to save parking charges or keep their vehicle close by. Unless that kind of empty-vehicle cruising is made illegal and enforced, the incentive for an owner to avoid it could be low, even if the streets are heavily congested.

Since VMT is correlated with vehicular wear and turnover, robotic vehicles in shared service would require an accelerated replacement rate compared to private, single-owner vehicles—at least four times faster to match a four-fold increase in PMT per vehicle. This could be made even worse depending on EMT rates. Hence, automotive production rates per VMT would be very unlikely to drop. Since the embedded carbon footprint of vehicle manufacture matches the fuel footprint of the first 120,000 miles or so of an ICE vehicle [Bern-11], accelerating the replacement rate of vehicles—including EVs—implies an expansion of CO2 footprint. In fact, if an EV were to run on 100% clean energy, then its footprint would be entirely from the embedded CO2 of manufacture and maintenance rather than sharing 50% of the blame with tailpipe emissions. This would tend to redirect GHG advocates’ attention toward the manufacture and disposal of vehicles and away from emissions concerns.

Still, it is true that vehicles in the 2040s could be smaller and made of recyclable graphene and numerous other factors that would shrink their manufacturing and operating footprint, but this effect remain to be seen. We still need to acknowledge that a lowered extant vehicle count is not a lowered vehicle manufacture count, because of higher VMT per vehicle. Rather it is only a lowered parking count—in itself critically important. Reduced congestion and fewer accidents from more intelligent driving are somewhat independent factors, although these should indirectly influence vehicle replacement rates as might a number of other factors. Consider, for example, accelerated vehicle wear and tear due to the tragedy of the commons associated with renters.

In a perfect world, we’d like to have the smallest number of vehicles possible provide all the PMT demanded with the minimum VMT required and with the minimum wait times for vehicle arrival. A major problem then becomes distributing fleets of shared vehicles in a way that provides the right type of vehicle (tailored) with a very short wait and minimal deadheading. The wrong vehicle and long waits encourages ownership, deadheading adds to VMT and congestion. Too many vehicles defeat the original goal. This turns out to be a difficult optimization problem—even without tailoring. It is made easier with large pools of users, i.e., users who do not own cars; hence the best way to ensure the success of massive sharing is… massive sharing. We will describe a way to address the tailoring problem using pricing in Chapter β.

 

Several factors will tend to increase PMT faster than the natural baseline set by the four socio-economic factors described above:

  1. It is easy to generate empty miles on chauffeuring trips that rely on robotic vehicles. Given the lack of time and effort needed to send a robotic vehicle on an errand, to pick up the drycleaning, send the family car to pickup an 11 year-old from scouts, go back home to park while at the office, or circle the block while its user shops, it will be easy to consume more EMT. Sending a vehicle for errands or chauffeuring does not affect the user’s Marchetti distance. This removes a critical barrier for EMT that would be in place for someone who previously had to drive for such purposes.
  2. Easier mobility for non-drivers and reluctant drivers. There are many reasons some people cannot or prefer not to drive. Robotic vehicles can remove many or most of these barriers. Some of this is currently handled by chauffeuring, but much is simply foregone. Easier access means more consumption of both PMT and VMT.
  3. Sprawling residential, commercial and retail. Many people prefer to live away from high-density areas. Office parks are not always mixed-use developments. Specialty retail destinations (e.g., Ikea, outlet shopping malls) are popular but have few locations with each targeting large geographic areas. Scale, selection and massive parking lots are their competitive edge.
  4. Last mile delivery. Last mile delivery will grow with more on-line shopping and more ordering-in as robotics lower the cost of meal access, convenient consumption and instant gratification. Trip chaining will move from a personal behaviour entirely into the domain of logistics operators. Last mile delivery may be an empty mile traveled or may require a human to handle or protect the goods being delivered. In the latter case, if a driver’s license is not required, hourly wages will be lower.
  5. Jevons paradox. If vehicles are privately owned, then smaller cheaper vehicles available to more owners would promote PMT. In the case of reliance on robo-fleets, cheaper miles and easier/quicker access to on-demand vehicles would drive up consumption of PMT.
  6. Media Richness Theory. The human preference for face-to-face meetings can more readily be met if the cost of mobility is reduced. Media Richness Theory says [Wiki-MRT] the interpersonal communication is richer (hence more effective) in person than via videoconference, phone or email. Hence, robo-transport might tend to replace telework or e-meetings to a degree.
  7. Easier trip-taking means shopping and entertainment facilities could become larger and more widely distributed continuing the existing trend toward fewer and more massive inventories and continuing to move the onus from logistics to extended travel and parking lots.
  8. The combination of a constant Marchetti budget and a decrease in congestion (due to robotics) would increase PMT.
  9. Private ownership is maintained. A large and growing vehicle population would increase parking demand which would increase parking congestion (circling) and PMT.
  10. Human preference for private space with occupancy controlled by the driver, including choice of traveling companions.

 

There several factors that would tend to dampen PMT growth:

  1. Congestion
  2. Non-responsive robotaxis (long service waits). This can also encourage ownership.
  3. High costs per mile, reduces the number of miles within one’s Marchetti budget.
  4. Higher urban density (reducing speed, increasing congestion), reduces the number of miles within one’s Marchetti distance.
  5. Vehicle taxes (policy).
  6. Road use charges (policy).
  7. Parking expense discourages travel by reducing the reach of one’s Marchetti budget.
  8. More walkable/bikeable communities increasing non-motorized substitution.
  9. Health awareness encourages spread of walk/bike.
  10. Robotic aviation for package delivery. Substitutes for EMT or for self-pickup (the latter motivated by acceptance checking, correctness of the order, does it fit, etc.)
  11. Concern for emissions. Assuming some fraction of non-renewable energy this might motivate fewer trips or more careful trip chaining for some users.
  12. Fear of using the road with robotic vehicles.

Caution: Points 8 and 9, in car-dependent societies, might merely exhibit high-percentage growth from a small, marginal base…

The fundamental human preference for automobility, combined with increased availability will mean that factors that tend to increase PMT growth would on the whole outweigh those that might tend to curtail it. The [Darg-07] wealth-ownership formulation showing that ownership saturates between 0.7 and 0.8, even if translated into equivalent PMT, will mean that, conservatively, motorized PMT levels will continue to grow to about four times current by 2050 and to about 8 times current as human population plateaus and its access to wealth continues to mature.

Hence, we emphasize that the need to satisfy this growth in PMT/EMT demand indicates that the choice between a shared-dominant and an ownership-dominant model for robotic vehicles must fall in favor of sharing. We further emphasize that the degree and methods of sharing current today are wholly inadequate to the requirement to be met by the 2040s.

Looking beyond the staging of vehicular robotics in the developed world, massive shared fleets will make it more feasible in most regions to satisfy ongoing PMT demands than would household ownership. The economics of this fact will do more to curtail worldwide ownership as robotics rush in to fill the PMT demand gap, than will the well meaning, liberal-American, sharing-economy perspective of shared-fleets as a way to reduce our burden on the environment. Most of the world wants less poverty more than they want clean air. And this is not about human populations growing in wealth, which will proceed apace. Rather it is about physical space for parked cars. There is simply not room in the significant cities in most countries that are now at 0.15 or 0.20 ownership levels as they aspire to PMT levels equivalent to ownership at 0.75.

So, robofleets will not be about saving the planet as much as they will be about addressing simple human demand for PMT and the fact that shared fleets can provide it far more quickly, more effectively and more cheaply than can household ownership. The figure below is a preview of what this already looks like today in a developing country.

traffic-transportation-lagos-nigeria_86774_990x742

Lagos, Nigeria sometime shortly before 2015. In 2007, vehicle population in Nigeria was at 31/1000. Image: http://ngm.nationalgeographic.com/2015/01/lagos/hammond-photography#/02a-vans-market-end-of-day-670.jpg

 

The AV is not really about U.S. millennials texting while traveling although that will be in the advertising paradigm for the developed world, or at least in North America. In the bigger picture, the AV is about better mobility for the other 4 billion people. The AV market in the developing world will far outperform that in the developed world.

What are some of the other implications? Countries like China and India will supply this market with cheap, serviceable vehicles. Compare the opportunity of supplying 500M vehicles to the market above or 1000 Mercedes-Benz F015s to rich Americans whose garages are already full. China is likely already planning for this. AI know-how is already spreading like a virus. Too many people are too smart and know too much. Today, AVs may look like an American or German or Swedish speciality, but that is an illusion given us by our blinkered media, and tech-hype magazines. In the half of the world that looks more like Lagos than Los Angeles, the hand wringing over insurance, AI decision ethics and most of rest of the current AV value-litany will pale in the face of this opportunity. Like water, technology runs downhill into the troughs and valleys of demand. Just as the developing world leapfrogged copper and went straight to cellular, it will leapfrog “a car in every garage” and go straight to robofleets.

We know wealth buys automobility, but we also know that mobility enables access to wealth. Regardless of whether that is a vicious or virtuous circle, the AV is capable of doing something far greater than is currently understood. Given the economics, it will be exploited. The AV has the potential to alter the economies of the rest of the Asia (as both manufacturers and consumers) and later Africa as a consumer and the labour pool to manufacture Chinese and Indian (etc) vehicles.

The sky-is-falling cries from transport infrastructure consultants pale. The save-the-planet-by-sharing crowd does not see past US cities. Shared fleets is a worldwide anti-poverty opportunity and should be forwarded at the United Nations level, rather than by the USDOT or by state DOTs.

 

[1] AV or Autonomous Vehicle is a fully automated vehicle or self-driving car that does not require a driver—or even a passenger. AVs are expected to be common well before mid century.

[2] VMT or vehicle miles traveled as shown on the vehicle odometer. VMT for the USA fluctuates around three trillion miles annually. Vehicle distance traveled is also expressed in VKT, vehicle kilometers traveled.

[3] PMT means personal miles traveled in a motor vehicle.  If two people travel in a car for ten miles, VMT=10 and PMT=20

[Bern-11] Berners-Lee, M., (2011) How Bad Are Bananas? The Carbon Footprint of Everything

[Cox-15] Cox, W., (2015) Behind the driving Increase. New Geography. http://www.newgeography.com/content/004874-behind-driving-increase

[Darg-07] Dargay, J., Gately, D., and Sommer, M., (2007) Vehicle Ownership and Income Growth, Worldwide: 1960-2030

[Grus-14] Grush, B., (2014) Social Evolution and road pricing. Tolling Review (Thinking Highways), 2014

[Marc-94] Marchetti, C. (1994) Anthropological Invariants in Travel Behaviour, Technological Forecasting and Social Change, 47, 75-88.

[Wiki-MRT] http://en.wikipedia.org/wiki/Media_richness_theory